INTERNATIONAL JOURNAL FOR NUMERICAL METHODS IN FLUIDS

Int. J. Numer. Meth. Fluids 2006; 52:11591174

Published online 7 April 2006 in Wiley InterScience (www.interscience.wiley.com). DOI: 10.1002/
d.1226

Application of the TVD scheme to the nonlinear

instability analysis of a capillary jet
Stephen G. Chuech1; ; and Ming-Ming Yan2;
1 Department

of Mechanical and Mechatronic Engineering; National Taiwan Ocean University;

Keelung 202; Taiwan
2 Chung-Shan Institute of Science and Technology; PO Box 90008-17-10; Lung-Tan;
Tao-Yuan 325; Taiwan

SUMMARY
In the past, when either the perturbation-type method or direct-simulation approach was used to analyse
capillary jets, the governing equations, which are parabolic in time and elliptic in space, were simplied
or linearized. In the present study, the convective derivative term and a full, nonlinear form of the
capillary pressure term are retained in the governing equations to investigate nonlinear eects on the
break-up of capillary jets. In this work, the TVD (i.e. total variation diminishing) scheme with
uxvector splitting is applied to obtain the solutions of the system of nonlinear equations in a matrix form.
Numerical results show that the present nonlinear model predicts longer jet break-up lengths and slower
growth rates for capillary jets than the previous linear model does. Comparing with other measurements
from past literatures, the nonlinear results are consistent with the experimental data and appear more
accurate than the linear analysis. In the past, the classic perturbation-type analyses assumed constant
growth rates for the fundamental and all harmonic components. By contrast, the present model is able
to capture the local features of growth rates, which are not spatially and temporally constant. Copyright
? 2006 John Wiley & Sons, Ltd.
KEY WORDS:

TVD scheme; capillary jet; jet break-up; growth rates; satellite drop

1. INTRODUCTION
An understanding of the liquid jet break-up process is of vital importance in the design of
many practical devices such as internal engine combustors [1, 2], gas turbines [3], liquid rocket
engines [4], and MEMS ejectors [57]. Applications for the capillary jet break-up range from
ink jet printing to drug delivery systems [57]. In ink jet printing, a capillary jet is ejected
Correspondence

to: Stephen G. Chuech, Department of Mechanical and Mechatronic Engineering, National Taiwan
Ocean University, Keelung 20224, Taiwan.
E-mail: sgc@mail.ntou.edu.tw
E-mail: yanmm@ms32.hinet.net

Copyright ? 2006 John Wiley & Sons, Ltd.

Received 25 September 2005

Revised 9 January 2006
Accepted 16 February 2006

1160

S. G. CHUECH AND M.-M. YAN

to produce small drops, which are often accompanied by satellite drops. The appearance of
satellite drops during the jet break-up process may result in print quality defects; therefore,
the issue of the capillary jet break-up still demands much attention.
Although the investigation of nonlinear eects on the jet break-up is critical to understand
the occurrence of satellite drops, previous studies [813] still utilized the linear
instability analysis by neglecting the convective derivative term in the momentum equation
due to the nonlinear complexity involved in the governing equations. In these linear analyses,
only the fundamental mode was considered and its growth rate was assumed to be constant.
In fact, the existence of harmonic modes other than the fundamental can be inferred from the
observation [14] that capillary jets do not break up into mono-size droplets, as the linear
theory would suggest.
Therefore, several studies, including perturbation-type analyses [1518] and directsimulation models [1927], preserved the convective derivative term to deal with the
nonlinear temporal instability. However, in the category of the perturbation type, their
analyses [1518] retained only the rst few harmonics in the perturbation expansions where
the wave growth rate was still assumed to be a constant of time. In the category of
direct methods, the one-dimensional models [19, 20] also utilized the expansion technique to
simplify the computational procedure. Lee [19] applied the LaxWendro method to the PDE
for numerical solutions in which the Taylor expansion was approximated only to the secondorder time step. Similarly, in Torpeys approach [20], the formation of satellite drops was
considered as a function of the amplitude of the secondary harmonic because the third and
higher harmonics were neglected. Other direct simulations [2127] circumvented the numerical diculties through linearization of the highly nonlinear capillary pressure term. The
full capillary pressure term, which does aect the satellite drop formation, should not be
simplied.
Although a full nonlinear transient NavierStokes system for the dynamics of drop formation was recently solved in two dimensions by Wilkes et al. [28], algorithms for obtaining
accurate solutions of the two-dimensional equations were computationally intensive due to the
required small time steps and ne grids across the entire jet and interface. Ambravaneswaran
et al. [29] remarked that one-dimensional models indeed hold certain advantages over their
two-dimensional counterparts, in terms often of both programming eorts and computational
time. Especially, it is questionable to immediately utilize the two-dimensional solutions to
obtain the jet break-up length. On the other hand, it should be noticed that all the one- or
two-dimensional analyses mentioned above focused only on the formation of main and satellite drops but reported no information of the jet break-up length and the growth rates of
disturbances on the jet surface.
Therefore, in this article, a one-dimensional system is adopted to study the instability and
break-up of capillary jets. In the governing equations, the convective derivative term and
the capillary pressure term in its full form are included to preserve all aspects of nonlinear nature. The equations in a matrix form, which are parabolic in time and elliptic
in space, will be solved by a TVD upwind scheme with
ux-vector splitting [30]. Meanwhile, the predictions for the jet break-up length and the wave growth rate are emphasized
in this study. For a validation of the present model, various experimental cases of Grant and
Middleman [31] will be simulated and compared for the jet break-up lengths. For a
further comparison, the local wave growth rates predicted by the present model are
averaged to compare with the linear solutions of Rayleigh [8] and experimental data of
Copyright ? 2006 John Wiley & Sons, Ltd.

Int. J. Numer. Meth. Fluids 2006; 52:11591174

NONLINEAR INSTABILITY ANALYSIS OF A CAPILLARY JET

1161

Donnely and Glaberson [32] though they should not be spatially and temporally constant.
The comparison results show that the nonlinear solutions for the break-up length and wave
growth rate are basically in agreement with the measurements. The nonlinear model predicts
longer break-up lengths and lower growth rates than the linear theory does.

2. GOVERNING EQUATIONS
One-dimensional study indeed holds advantages of large savings in programming and computational time. Therefore, the present paper adopts a one-dimensional model with the assumption
of plug jet velocity prole to study the instability and break-up of a capillary jet, as shown in
Figure 1. The present study is an application of the TVD scheme to the capillary jet breakup. One of main objectives of the article concerns whether the TVD scheme may handle the

ux-splitting phenomenon appearing inside the jet thread during the break-up process. Therefore, liquid viscosity and gravity were neglected for the study to facilitate the analysis of the
TVD scheme. In the model, it was assumed that the eects of turbulence, cavitation, velocity
prole, relaxation, and helical and KelvinHelmholtz instabilities played no role, so that only
the capillary instability appeared in the break-up process. The model also considered the
capillary jet as an innitely long cylinder without the aerodynamic eects of ambient air. The
liquid jet was assumed to be stationary relative to a moving observer. Thus, when the jet is
initially at rest to be perturbed by an innitesimal sinusoidal disturbance, the transient motion
of inviscid liquid will be governed by the following continuity and momentum equations:
@
1 @
(rur ) + (uz ) = 0
r @r
@z

(1)

@
1 @
@
1 @p
(uz ) +
(ruz ur ) + (uz2 ) =
@t
r @r
@z

@z

(2)

where uz and ur represent the axial and radial perturbation velocities, respectively,
is liquid
density, and p is the perturbation pressure.

Figure 1. Unstable disturbances growing on the surface of a capillary jet.

Copyright ? 2006 John Wiley & Sons, Ltd.

Int. J. Numer. Meth. Fluids 2006; 52:11591174

1162

S. G. CHUECH AND M.-M. YAN

By integrating Equations (1) and (2) with respect to r from the jet centreline to the jet
interface, the equations become
hur +

h2 @
(uz ) = 0
2 @z

h2 @
h2 @ 2
h2 @p
(uz ) + huz ur +
(uz ) =
2 @t
2 @z
2
@z

(3)
(4)

where h is the local jet radius h = a+; a and  denote the radius of the capillary tube and the
local amplitude of the growing disturbance on the jet surface, respectively. At the interface, 
is essentially the interfacial displacement function of t and z, and the radial velocity ur should
be equal to the total derivative of the (t; z). Therefore, the relationship between the radial
and axial velocities can be related to the local jet radius:
ur =

@h
@h
+ uz
@t
@z

(5)

Substituting Equation (5) into Equations (3) and (4), the equations [25] become
@ 2
@
(h ) + (h2 uz ) = 0
@t
@z
@ 2
@
h2 @p
(h uz ) + (h2 uz2 ) =
@t
@z

@z



 2 1=2
 2 3=2
1
2
@h
@h
@h
1+
2 1+
p = p = 
h

@z
@z
@z

(6)
(7)

(8)

Owing to surface tension  involved in the interfacial equilibrium for the capillary jet,
the pressure p should balance with the capillary pressure p [25, 33]. The rst term on the
right-hand side of Equation (8) is a destabilizing source, which results in the necking process
and leads to the satellite drop formation. On the other hand, the second term in Equation (8)
is a stabilizing source and always induces a restoring force.

3. NUMERICAL METHOD
To facilitate the analysis, Equations (6) and (7) are expressed in dimensionless form as
follows:
@v
@r
+
=0
@T
@Z

(9)

1
@ 2
@ 1
@v
+
(v = r)
= r (R 1 + R 2 )
(10)
@T
@Z
@Z

where dimensionless time T = t=
a3 = and dimensionless axial coordinate Z = z=a. The
variables appearing in Equations (9) and (10) are dimensionless groups: r = (h=a)2 = H 2 ,

Copyright ? 2006 John Wiley & Sons, Ltd.

Int. J. Numer. Meth. Fluids 2006; 52:11591174

NONLINEAR INSTABILITY ANALYSIS OF A CAPILLARY JET

1163


v = uz h2
=a3 , and R 1 and R 2 are mutually orthogonal principal radii of curvature of the jet
surface in dimensionless form:


2 1=2
@H

(11)
R1 = H 1 +
@Z


R 2 =

@2 H
@Z 2

1 

1+

@H
@Z

2 3=2

(12)

Furthermore, Equations (9) and (10) can be written in a matrix form as:
@G @E
+
=S
@T
@Z
where

(13)

 

G=


E=

S=

(14)

v
v

(15)

v2 = r
 
0

(16)

The capillary term in its full form in Equations (10)(12) was treated as a source term s
for the matrix system and no curve tting was used during the whole computation. A TVD
upwind scheme with
ux-vector splitting [30] was used to solve Equation (13). In the TVD
scheme, the
ux-vector E may be expressed directly in terms of its Jacobian A as E = AG.
Note that A is the Jacobian with respect to r and v.
Thus, we can split the
ux-vector into
two parts:
E+ =

(A + |A|)
G
2

(17)

E =

(A |A|)
G
2

(18)

where |A| = M ||M 1 , || is a diagonal matrix with absolute eigenvalues of A, and M 1
is an inverse of the matrix M which diagonalizes A. Hence, Equation (13) can now be
written as
@G @E + @E
+
+
=S
@T
@Z
@Z
Moreover, the equation can be rewritten in nite-dierence form as


k
k
Gik+1 = Gik Ei+
+ Sik T
1 E
1
i
2

Copyright ? 2006 John Wiley & Sons, Ltd.

(19)

(20)

Int. J. Numer. Meth. Fluids 2006; 52:11591174

1164

S. G. CHUECH AND M.-M. YAN

where T = (T k+1 T k ), = (T k+1 T k )=(Zi+(1=2) Zi(1=2) ), and a numerical
ux function

is introduced and dened as follows:
Ei(1=2) = 12 [A(Gi + Gi1 )] 12 |A|[Gi Gi1 ]

(21)

4. BOUNDARY AND INITIAL CONDITIONS

In the numerical computation, the disturbance wavelength  was taken as a computational
domain for a given dimensionless wavenumber, K = 2a=. The boundaries at the right and
left ends of this one-dimensional computational domain were prescribed with periodicity. In
other words, the calculation eld was kept xed over one wavelength with cyclic boundary
conditions. This is the feature of a standing wave for which the coordinate is moving with
the jet velocity as assumed previously. As a result, spatially periodic contracting and bulging
on the jet surface are developing with time marching. Initially, the local jet radius along the
z-axis direction at T = 0 can be expressed in dimensionless form as
h = 0 cos(kz) + a

(22)

where k is the wavenumber and 0 is the initial amplitude. Since the system is initially at
rest, the velocity is zero over the whole
ow eld at T = 0. Once the axisymmetric sinusoidal
disturbance is imposed on the jet surface, the curvature of the wave prole will induce a
capillary force to initiate the
ow motion.
The dependent variables were solved through the Equations (13)(21) with the boundary
and initial conditions mentioned above. A uniform one-dimensional grid system was adopted
for all computations. For the grid system, 50 computational cells were used for most cases
while a 100-cell grid system was employed for some cases of dimensionless wavenumber K
close to zero or one due to small growth rates or necessary resolutions for tiny satellites. In
order to avoid numerical instability, the step sizes of time marching were limited to 5% of
the quotient of the grid size divided by the characteristic velocity.

5. DROP FORMATION
The linear instability theory [8] shows that the cut-o dimensionless wavenumber is always
one for the capillary jet. In other words, when K is greater than one, the second term of the
capillary pressure in Equation (8) is a stabilizing source and will induce a restoring force.
As a result, the initial disturbance wave decays with time and the jet surface will eventually
restore to the originally undisturbed position. In order to simulate the surface wave evolution
and drop formation of an unstable jet in the break-up process, the value of K less than one
should be selected.
According to the linear theory, solutions of the dispersion relation providing the maximum
growth rate correspond to a value of the dimensionless wavenumber K approximately equal to
0.7. Therefore, the most unstable jet, perturbed at K = 0:7, was rst simulated for the evolution
in time of the jet surface wave. The
ow conditions for the experimental cases of Rutland
and Jameson [14] were adopted similar to the present test conditions. Figure 2 indicates that
the growth of the disturbance wave on the perturbed jet surface is not signicant at the early
Copyright ? 2006 John Wiley & Sons, Ltd.

Int. J. Numer. Meth. Fluids 2006; 52:11591174

NONLINEAR INSTABILITY ANALYSIS OF A CAPILLARY JET

2
1
0
-1
-2
-2
-1
0
H 1

1165

1
z /

Figure 2. Wave prole on the jet surface for K = 0:7 at T = 3:57.

2
1
0
-1
-2
-2
-1
0
H 12

1
z /

Figure 3. Wave prole on the jet surface for K = 0:7 at T = 10:28.

2
1
0
-1
-2
-2-1
H

01

z /

Figure 4. Wave prole on the jet surface for K = 0:7 at T = 13:05.

stage of T = 3:57. The results in Figures 3 and 4 show that a neck gradually forms at the
trough position and the trough contracts at a faster rate than the swell grows from T = 10:28 to
13.05. This phenomenon gives evidence that the wave growth rate is not constant spatially and
temporally. Figure 5 shows that the jet is about to disintegrate at T = 15:47. From Figure 5,
Copyright ? 2006 John Wiley & Sons, Ltd.

Int. J. Numer. Meth. Fluids 2006; 52:11591174

1166

S. G. CHUECH AND M.-M. YAN

2
1
0
-1
-2
-2
-1

2
0

1
z /

Figure 5. Formation of main and satellite drops before the jet break-up for K = 0:7 at T = 15:47.

2
H

1
0
-1
-2
-2
-1
0
H 12

1
z/

Figure 6. Formation of main and satellite drops before the jet break-up for K = 0:9 at T = 18:67.

it can be seen that as soon as the jet surfaces at two ends of the slender neck touch the
centreline, the swell portions will become main drops and the neck will turn into a satellite
drop in between at the trough position.
The evolution proles of the jet surface wave can be further demonstrated by the cases of
K = 0:9 and 0.3, as shown in Figures 6 and 7. Figure 6 shows the wave prole for K = 0:9
just before the jet break-up at T = 18:67. Comparing to the case of K = 0:7, the trough still
contracts faster than the swell grows, but the slender neck becomes shorter to form a smaller
satellite drop due to the short wavelength of K = 0:9. It is obvious that the case of K = 0:9
takes longer to break up because the disturbance wave of K = 0:7 grows at the fastest rate
according to the linear analysis.
For the case of K = 0:3, where the disturbance possesses a relatively long wavelength, the
trough contracts at a slower rate than the swell grows, and eventually produces a relatively
large satellite drop, as shown in Figure 7. This phenomenon is contrary to the cases of K = 0:7
and 0.9 with relatively short wavelengths. From Figures 57, it can be observed that the size
of satellite drop is basically smaller than that of main drop in the break-up process. However,
the size of satellite drop increases with increasing the disturbance wavelength and gradually
becomes comparable to that of main drop.
Copyright ? 2006 John Wiley & Sons, Ltd.

Int. J. Numer. Meth. Fluids 2006; 52:11591174

1167

3
2
1
0
-1
-2
-3
-3
-2
-1
H 01

NONLINEAR INSTABILITY ANALYSIS OF A CAPILLARY JET

2
2

1
z /

Figure 7. Formation of main and satellite drops before the jet break-up for K = 0:3 at T = 27:18.

6. JET SHAPE AND BREAK-UP LENGTH

The axisymmetric disturbance of K1 will exponentially grow and propagate along the
ow
direction to disintegrate the jet when the amplitude becomes comparable to the undisturbed jet
radius. To simulate such a jet shape, a water cylinder of diameter d = 0:4 mm, issuing from a
nozzle at a velocity of U = 1:5 m=s, was rst considered for K = 0:33, 0.48, 0.6, 0.7, 0.8, and
0.9. During the computations, the evolution of the wave prole was traced and plotted along
the jet surface according to
ow velocity. It can be seen in Figure 8 that the wave proles
on the jet surface are not signicant in the upstream region, but a satellite drop forms at the
jet tail. Similar to the results in Figures 57, the size of satellite drop is comparable to that
of main drop for the case of K = 0:33. On the contrary, the satellite drop becomes smaller
as the disturbance wavelength decreases. In Figure 8, the jet of K = 0:7 exhibits the shortest
length owing to its characteristics of the fastest growth.
Rutland and Jameson [34] observed the jet shape using a series of water jets perturbed
at various wavenumbers from K = 0:075 to 0.683. In the present study, two of their cases
(i.e. K = 0:25 and 0.683) were simulated and shown in Figure 9. Again, the size of satellite
drop is comparable to that of main drop for the long-wavelength case of K = 0:25. By contrast,
the jet of K = 0:683 exhibits a much shorter length and produces relatively smaller drops. The
present predicted jet shapes and lengths of K = 0:25 and 0.683 are in good agreement with
the photographs of Rutland and Jameson [34] in the literature. The only dierence is that in
the photographs [34] several main and satellite drops, which are separate in space, can be
observed beyond the jet tail. The present simulations in Figure 9 were performed only up to
the jet disintegration location where the jet surface contracts to touch the centreline.
A value for the continuous length L of the jet prior to break-up is termed break-up length.
In describing jet instability, one usually refers to the behaviour of the break-up length as a
function of the jet velocity. The break-up length of a liquid jet in a Rayleigh regime can
be obtained from the relation L = Ut. The linear relationship is reasonably well described
by Webers analysis [9] for the break-up of a laminar Newtonian jet under the in
uence of
surface tension forces alone. According to Webers theory, the jet is assumed to be subject
Copyright ? 2006 John Wiley & Sons, Ltd.

Int. J. Numer. Meth. Fluids 2006; 52:11591174

1168

S. G. CHUECH AND M.-M. YAN

Figure 8. Calculated jet shapes of a jet perturbed at various wavenumbers.

Figure 9. Numerical simulations of the experimental cases of Rutland and Jameson [34].

to a spectrum of disturbance of the Fourier form, but to suer disruption under the action of
the disturbance with the largest growth rate. Figure 10 shows the present predictions of water
jets of d = 0:4 mm issuing at various jet velocities, namely U = 1:25, 1.37, 1.50, 1.57, and
1:72 m=s. In the analysis, these jets were assumed to be perturbed by a disturbance of K = 0:7,
which should grow the fastest. From Figure 10, it can be clearly seen that the changes of
the velocity apparently do not aect the sizes of main and satellite drops. The results show
that the break-up length increases with the jet velocity, and the relationship between them is
basically linear.
Copyright ? 2006 John Wiley & Sons, Ltd.

Int. J. Numer. Meth. Fluids 2006; 52:11591174

NONLINEAR INSTABILITY ANALYSIS OF A CAPILLARY JET

1169

Figure 10. Numerical simulations of water jets issuing at various velocities.

Figure 11. Numerical simulations of water jets issuing from various nozzles.

Besides the velocity, the jet diameter is also an essential factor to aect the jet behaviour.
Figure 11 shows the eects of the diameter on drop size and break-up length using the case
of U = 1:5 m=s in Figure 10, but with three dierent diameters of d = 0:31, 0.40, 0.62 mm.
It is obvious that the larger the jet diameter is, the bigger the drops are produced, as shown
in Figure 11. Meanwhile, the break-up length also increases nonlinearly with increasing the
diameter size of the jet. Furthermore, the values of L=d for the cases of d = 0:31, 0.40,
0.62 mm (i.e. 45.56, 53.04, and 66.50, respectively) are nonlinearly proportional to the Weber
Copyright ? 2006 John Wiley & Sons, Ltd.

Int. J. Numer. Meth. Fluids 2006; 52:11591174

1170

S. G. CHUECH AND M.-M. YAN

Figure 12. Variations of normalized break-up length with Weber number.

numbers of these cases (i.e. 9.795, 12.64, and 19.59, respectively). Note that the denition of
the Weber number is We =
dU 2 =.
Among a spectrum of disturbance, the most unstable wave corresponding to K = 0:7 should
rst satises the break-up condition. Figure 12 shows that the present prediction for the breakup lengths of capillary jets subjected to a disturbance of K = 0:7 was compared to experimental
work of Grant and Middleman [31], which is an empirical correlation of ve dierent liquid
jets with a variety of diameters. The linear theory of Levich [33] was also applied to predict
the jet break-up length. As shown in Figure 12, the curve of the present prediction appears
nonlinear in the range of Weber number from 0 to 40 while it becomes approximately linear
for higher Weber numbers. Note that the nonlinear behaviour of the break-up length for lower
Weber numbers is similar to the results of the cases in Figure 11. The prediction of the linear
theory of Levich [33] exhibits the similar behaviour along the curve, but it is underestimated.
Based on the same
ow conditions, the curve predicted by the present nonlinear model shows
a better agreement with the measurements than the linear analysis. The reason for this may
be that the present simulation can account for the existence of all harmonics and their growth
variations.
7. GROWTH RATE OF SURFACE WAVE
The variations of logarithmic value of the wave amplitude predicted by the linear theory of
Rayleigh [8] and the nonlinear theory of Yuen [15] were linear with time. In other words, the
wave growth rates for the fundamental and harmonics modes were all assumed to be constant
in their analyses. In fact, it has been observed by Goedde and Yuen [35] that for an actual
liquid jet the surface growth at each location along the wavelength is dierent and varies with
time. Goedde and Yuen [35] also plotted the logarithmic value of the dierence between the
amplitude at the swell and that at the neck, and then used the linear region on the curve to
Copyright ? 2006 John Wiley & Sons, Ltd.

Int. J. Numer. Meth. Fluids 2006; 52:11591174

NONLINEAR INSTABILITY ANALYSIS OF A CAPILLARY JET

1171

Figure 13. Variations of local growth rate with time.

describe the growth rate. Later, Ashgriz and Mashayek [36] utilized the similar time-averaging
process to examine the temporal growth of the disturbances on viscous jet surface. However,
it should be pointed out that a single, time-averaging growth rate does not properly describe
the actual behaviour of the jet.
Therefore, a formulation for the instantaneous growth rate at any location along the wavelength is proposed in the present study as follows:
Ai =

=Ki )
ln(K+1
i
T K+1 T K

(23)

where i is the dimensionless amplitude of the disturbance at t he location i along the wavelength and Ai represents the instantaneous, local growth rate. Figure 13 presents the variations
of the local growth rates at the swell and neck points with time for the disturbance of K = 0:7.
For this case, the growth rates at both locations increase quickly from T = 0 to 10, and then
they keeps almost constant until the jet is about to break up. It is interesting that the
experimental work of Goedde and Yuen [35] showed the similar results. They reported that
the logarithmic value of the dierence between the amplitude at the swell and that at the
neck varies linearly except close to the break-up time. As seen in Figure 13, the present
analysis demonstrates that the growth rate of the swell increases suddenly and rapidly, but,
on the contrary, the neck stops to grow just before main and satellite drops are detached from
the jet.
In order to compare the present results with other data in the previous literatures, an
average value of the constant regions on the curves in Figure 13 was adopted to represent the
global growth rate of the disturbance, even though it is not proper to describe the instability
of the jet with an average growth rate. Figure 14 shows the global growth rate A as a
function of wavenumber K; for the comparison, the experimental results of Donnely and
Glaberson [32] and the linear solutions of Rayleigh [8] are also shown. These results show
that for the long waves (K1) the growth rate is positive, and hence the long waves can grow;
Copyright ? 2006 John Wiley & Sons, Ltd.

Int. J. Numer. Meth. Fluids 2006; 52:11591174

1172

S. G. CHUECH AND M.-M. YAN

Figure 14. Variations of global growth rate with wavenumber.

for the short waves (K1), the growth rate is zero, and therefore the shore waves do not
grow. As shown in Figure 14, the present nonlinear model predicts slightly lower growth rates
than those obtained from the dispersion equation of Rayleigh [8] though both predictions are
in agreement with the measurements. For K0:6, the present nonlinear predictions basically
agree better with the measurements than the solutions of the linear theory do.

8. CONCLUSIONS
Major conclusions of the present study can be summarized as follows.
1. The sizes of main and satellite drops produced from capillary jets basically increase
with increasing disturbance wavelength or nozzle diameter. The change of jet velocity
apparently does not aect the sizes of the drops as long as the nozzle diameter keeps
the same and the jet is perturbed at a xed wavenumber.
2. The present results show that break-up length increases nonlinearly with increasing nozzle
diameter. The results also conrm the Webers linear relationship [9] between break-up
length and jet velocity. However, when the break-up length is normalized by the nozzle
diameter, the normalized break-up length appears nonlinear for low Weber numbers, and
it gradually becomes linear as the Weber number increases.
3. In the comparison study, both linear and nonlinear predictions for break-up length are
consistent with the distribution of experimental data, but the linear solutions are slightly
underestimated. The reason for this may be attributed to the fact that the linear theory
applies only the fundamental with a constant growth rate, but the present nonlinear
simulation can account for the existence of all harmonics and their growth variations.
However, a further study to validate this attribution is needed.
Copyright ? 2006 John Wiley & Sons, Ltd.

Int. J. Numer. Meth. Fluids 2006; 52:11591174

NONLINEAR INSTABILITY ANALYSIS OF A CAPILLARY JET

1173

4. The global growth rates predicted by the present nonlinear model are slightly lower than
those obtained from the dispersion equation of the linear theory. It is coincidental that
a slower growth rate gives a longer jet break-up length. In fact, a single value of the
global growth rate should not be used to describe the nonlinear behaviour of the jet. In
the present study, local growth rates are computed, and the results show that the local
growth rates for the swell and the neck increase with time at the early stage, and then
they keep almost unvarying until the break-up occurs. Therefore, it is evident that the
disturbance growth indeed varies temporarily though the wave grows steadily most of
time.
5. The present computations show that the trough contracts at a slower rate than the swell
grows for the case of long wavelength (e.g. K =0:3). By contrast, the trough contracts
at a faster rate than the swell grows for the case of short wavelength (e.g. K = 0:7 and
0.9). This phenomenon implies that the growth rate is also not constant in space so that
local growth rates along the wavelength should be emphasized in the instability analysis
of a capillary jet.

REFERENCES
1. Lefebvre AH. Atomization and Sprays. Hemisphere: New York, NY, 1989.
2. Lin SP, Reitz RD. Drop and spray formation from a liquid jet. Annual of Review Fluid Mechanics 1998; 30:
85105.
3. Lefebvre AH. Gas Turbine Combustion. Hemisphere: New York, NY, 1989.
4. Inamura T, Daikoku M. Numerical simulation of droplet formation from coaxial twin-
uid atomizer. Atomization
and Sprays 2002; 12:247266.
5. Gooray A, Roller G, Galambos P, Zavadil K, Givler R, Peter F, Crowley J. Design of a MEMS ejector for
printing applications. Journal of Imaging Science and Technology 2002; 46:415421.
6. Tseng FG, Kim CJ, Ho CM. Microinjector Free of Satellite Drops and Characterization of the Ejected
Droplets, vol. 66. ASME, DSC Division MEMS (Publication): New York, 1998; 8995.
7. Cheng S, Chandra S. A pneumatic droplet-on-demand generator. Experiments of Fluids 2003; 34:755762.
8. Rayleigh L. Theory of Sound. Macmillan: London, 1896 (Dover, New York, reprinted in 1945).
9. Weber C. Zum zerfall eines
ussigkeitsstrahles (on the disruption of liquid jets). Zeitschrift fur Angewandte
Mathematik und Mechanik 1931; 2:136141.
10. Sterling AH, Sleicher CA. The instability of capillary jets. Journal of Fluid Mechanics 1975; 68:447495.
11. Reitz RD, Bracco FV. Mechanism of atomization of a liquid jet. Physics of Fluids 1982; 25:17301742.
12. Lin SP, Kang DJ. Atomization of a liquid jet. Physics of Fluids 1987; 30:20002006.
13. Chuech SG, Przekwas AJ, Singhal AK. Numerical modeling for primary atomization of liquid jets. AIAA
Journal of Propulsion and Power 1991; 7:879886.
14. Rutland DF, Jameson GJ. A non-linear eect in the capillary instability of liquid jets. Journal of Fluid Mechanics
1971; 46:267272.
15. Yuen MC. Non-linear capillary instability of a liquid jet. Journal of Fluid Mechanics 1968; 33:151163.
16. Lafrance P. Nonlinear breakup of a laminar liquid jet. Physics of Fluids 1975; 18:428432.
17. Chaudhar KC, Redeko LG. The nonlinear capillary instability of a liquid jet: Part 1 theory. Journal of Fluid
Mechanics 1980; 96:257274.
18. Bogy DB. Steady draw-down of a liquid jet under surface tension and gravity. Journal of Fluid Mechanics
1981; 105:157176.
19. Lee HC. Drop formation in a liquid jet. IBM Journal of Research Development 1974; 18:364369.
20. Torpey PA. A nonlinear theory for describing the propagation of disturbances on a capillary jet. Physics of
Fluids A 1989; 1:661671.
21. Sellens RW. A one-dimensional numerical model of capillary instability. Atomization and Sprays 1992; 2:
239251.
22. Shokoohi F, Elrod HG. Numerical investigation of the disintegration of liquid jets. Journal of Computational
Physics 1987; 71:324342.
23. Childs RE, Mansour NN. Simulation of fundamental atomization mechanism in fuel sprays. AIAA 26th
Aerospace Science Meeting 1988, January 1114, Reno, Nevada, Paper no. 88-0238.
Copyright ? 2006 John Wiley & Sons, Ltd.

Int. J. Numer. Meth. Fluids 2006; 52:11591174

1174

S. G. CHUECH AND M.-M. YAN

24. Fromm JE. Numerical calculation of the
uid dynamics of drop-on-demand jets. IBM Journal of Research
Development 1984; 28:322333.
25. Eggers J, Dupont TF. Drop formation in a one-dimensional approximation of the NavierStokes equation.
Journal of Fluid Mechanics 1994; 262:205221.
26. Eggers J. Nonlinear dynamics and breakup of free-surface
ows. Reviews of Modern Physics 1997; 69:865929.
27. Brenner MP, Eggers J, Joseph K, Nagel SR, Shi XD. Breakdown of scaling in droplet ssion at high Reynolds
number. Physics of Fluids 1997; 9:15731590.
28. Wilkes ED, Phillips SD, Basaran OA. Computational and experimental analysis of dynamics of drop formation.
Physics of Fluids 1999; 11:35773598.
29. Ambravaneswaran B, Wilkes ED, Basaran OA. Drop formation from a capillary tube: comparison of onedimensional and two-dimensional analyses and occurrence of satellite drops. Physics of Fluids 2002; 14:
26062621.
30. Anderson JD. Computational Fluid Dynamics. McGraw-Hill: New York, NY, 1995.
31. Grant RP, Middleman S. Newtonian jet stability. AIChE Journal 1966; 12:669678.
32. Donnely RJ, Glaberson W. Experiment on the capillary instability of a liquid jet. Proceedings of the Royal
Society of London, Series A 1965; 290:547556.
33. Levich VG. Physicochemical Hydrodynamics. Prentice-Hall: New Jersey, NJ, 1962.
34. Rutland DF, Jameson GJ. Theoretical prediction of the sizes of drops formed in the breakup of capillary jets.
Chemical Engineering Science 1970; 25:16891968.
35. Goedde EF, Yuen MC. Experiments on liquid jet instability. Journal of Fluid Mechanics 1970; 40:495511.
36. Ashgriz N, Mashayek F. Temporal analysis of capillary jet breakup. Journal of Fluid Mechanics 1995; 291:
163190.

Copyright ? 2006 John Wiley & Sons, Ltd.

Int. J. Numer. Meth. Fluids 2006; 52:11591174